Tiantian Gong

Tiantian Gong, Guodong Du, Junsheng Wang et al.

Traditional text-based person re-identification (ReID) techniques heavily rely on fully matched multi-modal data, which is an ideal scenario. However, due to inevitable data missing and corruption during the collection and processing of cross-modal data, the incomplete data issue is usually met in real-world applications. Therefore, we consider a more practical task termed the incomplete text-based ReID task, where person images and text descriptions are not completely matched and contain partially missing modality data. To this end, we propose a novel Prototype-guided Cross-modal Completion and Alignment (PCCA) framework to handle the aforementioned issues for incomplete text-based ReID. Specifically, we cannot directly retrieve person images based on a text query on missing modality data. Therefore, we propose the cross-modal nearest neighbor construction strategy for missing data by computing the cross-modal similarity between existing images and texts, which provides key guidance for the completion of missing modal features. Furthermore, to efficiently complete the missing modal features, we construct the relation graphs with the aforementioned cross-modal nearest neighbor sets of missing modal data and the corresponding prototypes, which can further enhance the generated missing modal features. Additionally, for tighter fine-grained alignment between images and texts, we raise a prototype-aware cross-modal alignment loss that can effectively reduce the modality heterogeneity gap for better fine-grained alignment in common space. Extensive experimental results on several benchmarks with different missing ratios amply demonstrate that our method can consistently outperform state-of-the-art text-image ReID approaches.

Tiantian Gong, Ryan Henry, Alexandros Psomas et al.

A long line of research on secure computation has confirmed that anything that can be computed, can be computed securely using a set of non-colluding parties. Indeed, this non-collusion assumption makes a number of problems solvable, as well as reduces overheads and bypasses computational hardness results, and it is pervasive across different privacy-enhancing technologies. However, it remains highly susceptible to covert, undetectable collusion among computing parties. This work stems from an observation that if the number of available computing parties is much higher than the number of parties required to perform a secure computation task, collusion attempts in privacy-preserving computations could be deterred. We focus on the prominent privacy-preserving computation task of multi-server $1$-private information retrieval (PIR) that inherently assumes no pair-wise collusion. For PIR application scenarios, such as those for blockchain light clients, where the available servers can be plentiful, a single server's deviating action is not tremendously beneficial to itself. We can make deviations undesired via small amounts of rewards and penalties, thus significantly raising the bar for collusion resistance. We design and implement a collusion mitigation mechanism on a public bulletin board with payment execution functions, considering only rational and malicious parties with no honest non-colluding servers. Privacy protection is offered for an extended period after the query executions.

Tiantian Gong, Mohsen Minaei, Wenhai Sun et al.

Mining processes of Bitcoin and similar cryptocurrencies are currently incentivized with voluntary transaction fees and fixed block rewards which will halve gradually to zero. In the setting where optional and arbitrary transaction fee becomes the remaining incentive, Carlsten et al.\ [CCS~2016] find that an undercutting attack can become the equilibrium strategy for miners. In undercutting, the attacker deliberately forks an existing chain by leaving wealthy transactions unclaimed to attract petty complaint miners to its fork. We observe that two simplifying assumptions in [CCS~2016] of fees arriving at fixed rates and miners collecting {\em all} accumulated fees regardless of block size limit are often infeasible in practice and find that they are inaccurately inflating the profitability of undercutting. Studying Bitcoin and Monero blockchain data, we find that the fees deliberately left out by an undercutter may not be attractive to other miners (hence to the attacker itself): the deliberately left out transactions may not fit into a new block without "squeezing out" some other to-be transactions, and thus claimable fees in the next round cannot be raised arbitrarily. This work views undercutting and shifting among chains rationally as mining strategies of rational miners. We model profitability of undercutting strategy with block size limit present, which bounds the claimable fees in a round and gives rise to a pending (cushion) transaction set. In the proposed model, we first identify the conditions necessary to make undercutting profitable. We then present an easy-to-deploy defense against undercutting by selectively assembling transactions into the new block to invalidate the identified conditions. Under a typical setting with undercutters present, applying this avoidance technique is a Nash Equilibrium. Finally, we complement the above analytical results with experiments.